Two key questions for understanding the visual system are, how is information about the world represented in the cortex, and how are these representations formed during development. The best studied cortical areas in this regard are V1 and V2, where features of the visual scene such as position, orientation, direction and spatial frequency are represented in maps with complex local and global structure. Understanding how these maps develop, and what principles underlie their final structure, is an important step in answering the above questions. The relative contribution of genetic versus epigenetic (particularly activity-dependent) mechanisms in determining visual cortical structure is still a subject of intense debate. This debate impacts directly on the design of effective therapies for treating the effects of early abnormal visual experience, such as strabismus. Visual cortical map structure arises from activity-dependent learning rules that attempt to represent highly correlated input features close together, acting in conjunction with genetically determined constraints such as the shape of the target area. Both the local and global structure of maps in V1 and V2 can be accounted for in terms of the interaction between the correlational structure of afferent activity, patterns of intracortical connections, and the shape of the cortical target region. Changes in map structure due to abnormal rearing paradigms such as monocular deprivation and strabismus follow naturally from the same rules. The hypothesis will be tested using computational models suited to exploring questions of large scale map structure. These models will be applied to low dimensional feature spaces representing position, orientation, ocular dominance, direction, spatial frequency, disparity, and color. Several types of parameter variation will be investigated, including variations in the statistical structure of afferent activity, variations in the pattern of intracortical connections, variations in the way similarity between input features is measured, and variations in the shape of the target region. At each stage, simulation results will be directly compared to experimental data using a range of quantitative techniques. The analysis and simulation of these computational models will lead to experimentally testable predictions concerning the parameters underlying cortical map formation and the effects of visual experience on cortical map structure.